AI Model Report

Model Releases · AUGUST 19, 2026

Gemini 3.7 Flash lands three weeks after 3.6, halves token price — and 3.5 Pro still has no date

Google shipped Gemini 3.7 Flash on August 13 with a 16-point DeepSWE v1.1 jump over its three-week-old predecessor, at $0.75/$3.75 per million tokens — half of 3.6's launch price. The flagship Pro remains missing.

By Karl Strauchman · Senior model reviewer · August 19, 2026

Google shipped Gemini 3.7 Flash on August 13, three weeks after 3.6 Flash, with a 16.3-point jump on DeepSWE v1.1 (65.3% versus 49.0%) and introductory API pricing of $0.75 per million input tokens and $3.75 per million output tokens, half of what 3.6 cost at launch. The workhorse line is moving fast. The flagship Pro isn't.

Bloomberg, Reuters, and VentureBeat all led with the same absence: Thursday's launch shipped no updated timetable for Gemini 3.5 Pro, the frontier model that missed its original target after falling short of internal coding goals, per Reuters reporting in July. Google still describes 3.5 Pro as delayed. SemiAnalysis has told Bloomberg the company effectively canceled it, a claim Google hasn't confirmed. Meanwhile, Reuters says training has begun on Gemini 4, described as Google's most ambitious model yet.

The numbers on 3.7 Flash are real. FrontierCode 1.1 Main climbs from 34.4% to 43.6%. Google attributes the gains to "algorithmic innovations" applied on top of the existing base rather than a fresh pretraining run, and VentureBeat reports the techniques will inform future models. The model ships as gemini-3.7-flash in the API and AI Studio, powers Google Antigravity, and runs Gemini Spark, the company's personal-agent product, in more than 160 countries. Introductory pricing holds through December 31, 2026, then doubles to $1.50/$7.50 per million tokens on January 1, 2027.

Read the release cadence and the price cut together and the strategy is legible. Google is compounding gains on the tier it can actually ship, at aggressive economics, while the tier it can't ship recedes into rumor. Flash is doing frontier work in press-release terms because Pro isn't there to do it.

The organizational backdrop makes the pattern harder to dismiss. DeepMind's leadership overhaul was announced last week, with Demis Hassabis stepping aside in favor of deputy Koray Kavukcuoglu. The two original Gemini technical co-leads have left, Noam Shazeer to OpenAI and John Jumper to Anthropic, to co-found a startup. Sergey Brin has reportedly urged staff to go "all in" on Gemini. That's the vocabulary of a company that knows it's behind on the metric that matters and is racing to compress the gap.

Three weeks between Flash releases is unusual. It suggests Google has algorithmic headroom on the smaller line and needs to be seen shipping. What it doesn't suggest is a Pro model close to shipping.

Sources